Central processing units (CPUs) are commonly used in modern robotics. They excel at quickly processing small streams of data, although they are limited by their inability to do more than one thing at a time. This can make them an inefficient choice for some tasks.
Neurala, a robotics and neuroscience start-up from Cambridge, Massachusetts, has experimented with the use of graphics processing units (GPUs) in their machines. It is hoped that this development will allow for the creation of robotic ‘brains’ that analyse information more like a human brain. GPUs have a better capacity for processing large sets of data, and are able to handle several at a time. This is closer to the way the human brain handles information.
GPUs have previously been too large and expensive to be considered for use in robotics; they are more commonly used for gaming and supercomputers. However, their recent use by Neurala has produced much faster machines.
These new robotic brains could potentially make space exploration easier. They could be used to control a rover, for example. The GPU allows a robot to process visual information (like the changing terrain found on planets like Mars) in real time. With this technology comes the possibility that rovers could be permitted to make their own decisions when exploring.
Neurala’s aim to mimic a human thought process has drawn attention from NASA. Mark Motter (of NASA’s Langley Research Centre) says “This is an interesting approach to autonomy.” The use of GPUs could prove to be the difference between automation and autonomy. Instead of an automated machine, which has a predetermined plan to execute, an autonomous robot could reach its goal by making independent decisions.
In addition to the ability to make choices for itself, Neurala’s GPU brain could have other benefits. Being able to process visual information could allow a rover to perform some simple scientific tasks. It could take note of any unusual formations that may require additional investigation, or it could be used to search for signals indicating the presence of water or minerals.
NASA’s most technologically advanced rover, Curiosity, is currently controlled exclusively by human operators. Because it is such an expensive machine, every precaution must be taken to ensure its continued operation. Neurala’s CEO, Max Versace, commented: “If I’m going to send a rocket that costs billions of dollars, I want to be sure that every millimetre I travel is gonna be safe”. However, as the cost of space flights decreases, Versace has noted the potential for larger numbers of inexpensive robots. “We need similar machinery to the biological brain to do that,” he says.
Neurala’s GPU brain could also serve to improve unmanned aerial vehicles (UAVs). If these drones were ever to share airspace with manned planes (they are currently only permitted to fly under 400 feet) they would be required to be more autonomous. The Federal Aviation Authority (FAA) has created several UAV testing sites in the US. It is hoped that the research carried out at these sites will eventually allow drones to sense and avoid other aircraft by themselves. They could then be flown throughout national airspace.
The GPU could also be developed alongside advances in mobile technology, according to Versace. He says that the hardware that smart phones are making available could be used to create mobile autonomous robots within a few years. Neurala have made progress into developing compatibility between GPUs and current mobile devices, like the iPad – “We are designing technology for Mars, [however] our goal is to bring it back to Earth.”
What other innovative applications might the GPU see in the future?